Evolving control for modular robotic units

Abstract

This paper describes a set of experiments aimed at exploring the use of artificial evolution to automatically produce controllers for a self-reconfiguring robot. The experiments are based upon initial experiments in programming distributed controllers in a simulation of the M-TRAN modules. This leads to a series of experiments that evolves and co-evolves controllers in a simulation of a new self-reconfigurable modular robot, called the ATRON. The results are discussed, and it is concluded that artificial evolution may be an approach for producing controllers for the individual ATRON module in such a way, that the collective of modules have a meaningful behavior.